ERA5-land conflicting expver

Software and Tools API
, ,
1

at the time of writing, the following request fails to produce a readable netcdf file:

{
  "variable": [
    "2m_temperature",
    "total_precipitation",
    "surface_solar_radiation_downwards",
    "2m_dewpoint_temperature",
    "10m_u_component_of_wind",
    "10m_v_component_of_wind",
    "soil_temperature_level_1",
    "soil_temperature_level_2"
  ],
  "date": [
    "2025-09-01"
  ],
  "time": [
    "00:00"
  ],
  "data_format": "netcdf",
  "download_format": "unarchived",
  "area": [
    61,
    -8.75,
    49.75,
    2.0
  ],
  "nocache": "4812921339199228"
}

A warning is logged by the python cdsapi client

WARNING Structural differences in grib fields detected when opening in xarray. Opening the grib file safely, however this may result in files with non-intuitive filenames.

Splitting up the variables into separate requests produces readable outputs, with the difference being that the soil variables have expver 0005, whilst the other variables have expver 0001.

It seems that any request for these variables that includes the above timestep will currently fail.

Including "expver": "0005" in the request seems to have no effect.

Is there a proper way to handle this?

1 Like
2

Cross-post: Different `expver` version for a single timestamp in day ERA5

It seems that the non-soil variables have this behaviour.
For date 2025-09-01, expver = 0001 @ step 00:00, and expver = 0005 @ step 01:00.

But not the soil variables, which have consistent expver = 0005